Curriculum/meta-labelling
Meta-Labelling
research process·L2 · idiom·stub
Replacesthe belief that one ML model should predict both direction and confidence.
Meta-labelling (López de Prado, AFML ch. 3) splits the prediction into two models: the primary model predicts direction (long / short / pass); the secondary model predicts whether to take the trade. Each model sees a different subset of features and labels. Precision-recall trade-offs improve at every operating point; false positives drop without sacrificing recall on confident trades.
Prerequisites
Unlocks—
Bridges
- primary-vs-secondary-model-decompositionmodel to implementationThe two-model split generalises to risk control (primary picks direction, secondary picks size) and to ensemble methods (primary is the base learner, secondary is the gate). The pattern is decomposition by *decision type*, not feature subset.
This concept is a node in the curriculum DAG. The full lab — page blocks, done state, references — has not been authored yet. The relations above describe where it sits in the graph.
Author at: content/concepts/meta-labelling/card.ts